3,023 Squirrels, One Park, One Census (Live NYC Open Data)

In October 2018, several hundred volunteers walked Central Park with clipboards and counted squirrels. Not estimated, not modelled, actually counted, one animal at a time, noting the color of each one, what it was doing, and whether it cared that a human was watching. The result is one of the most charming open datasets in existence: 3,023 individual squirrel sightings, each with dozens of fields. We loaded the whole thing into a public dataset on QuantumLayers to see what a single autumn snapshot of one park’s squirrels can tell us.

The dataset is marked public, and every chart below is drawn live: when this page loads, your browser calls QuantumLayers’ public chart endpoint and renders whatever the dataset currently holds, with no API key, no login, and no hard-coded numbers. Each figure also links to its own standalone ql-chart page on QuantumLayers.

The whole park, drawn by its squirrels

Start with the simplest thing you can do with 3,023 pairs of coordinates: plot them. Every dot below is one squirrel, placed where it was seen and colored by its fur. You are not looking at a map of Central Park with squirrels added; you are looking at the park’s footprint reconstructed entirely from where squirrels happened to be standing. The long green rectangle is the park itself, traced out by its most abundant residents.

Figure 1: Every one of the 3,023 sightings by location, colored by primary fur color, fetched live from QuantumLayers. Source: NYC Open Data.

▶ Open the live ql-chart on QuantumLayers

Are they really all gray?

The Eastern gray squirrel is named for the obvious reason, and the data agrees emphatically. Of the squirrels whose coat was recorded, 83% were gray. But the park is not a monoculture: about 13% were cinnamon, a warm reddish-brown morph of the same species, and a rarer 3.5% were black, the melanistic variant that turns heads whenever it appears. Same species, three very different outfits.

Figure 2: Share of squirrels by primary fur color, fetched live. “NA” is sightings where color was not recorded. Source: NYC Open Data.

▶ Open the live ql-chart on QuantumLayers

Up a tree, or down in the grass?

For every sighting, counters noted whether the squirrel was on the ground or somewhere above it. The ground wins by a wide margin: about 72% of squirrels were at ground level, foraging, sitting, or dashing across a path, while the remaining 28% were up a tree, on a bench, or on some other raised surface. When they did climb, they rarely went far: among sightings where a height was recorded, the median was just 10 feet, though a few daredevils were logged as high as 180 feet up.

Figure 3: Ground level versus above ground, fetched live. Source: NYC Open Data.

▶ Open the live ql-chart on QuantumLayers

What the statistics reveal

Charts show the shape; the statistics pin it down. We ran QuantumLayers’ automated analysis across the categorical columns, testing which behaviors and traits actually predict each other rather than just co-occurring by chance. Three findings stood out.

A squirrel is either climbing or foraging, almost never both. The single strongest relationship in the whole behavior dataset is between climbing and foraging (Cramér’s V ≈ 0.30, χ² ≈ 268, p < 0.0001). That makes intuitive sense once you picture it: foraging is a nose-down, ground-level activity, and climbing is the opposite posture entirely. The data has cleanly captured two mutually exclusive modes of being a squirrel. Running and chasing show the same tension with foraging, just more weakly.

Grown-ups get down to business. Age is significantly associated with foraging (χ² ≈ 30, p < 0.0001). Adults were recorded foraging in about half of their sightings, while juveniles foraged in barely a third. Flip it around and the young ones are more often just playing: juveniles turned up climbing and eating at higher rates than adults. The serious food-gathering is grown-up work; the youngsters are still messing about.

Fur color tracks with where you find them. QuantumLayers flagged a significant association between primary fur color and location in the park (χ² ≈ 220, p < 0.0001). This is almost certainly not about the squirrels and everything about the park: different areas have different tree cover, foot traffic, and counting conditions, and the color morphs are not spread perfectly evenly across them. It is a reminder that even a squirrel census is, in the end, a map of habitat and of the humans doing the counting.

▶ Open the live ql-chart on QuantumLayers

How this was built

No spreadsheets, no manual downloads. The Squirrel Census file was connected as a public dataset in QuantumLayers, which ingested and typed every column. Each chart above is rendered with Chart.js, populated at page load by a single POST to QuantumLayers’ ql_get_chart_data endpoint. Because the dataset is public, that call needs no authentication: the browser gets back a ready-to-draw Chart.js config and paints it. The “Open the live ql-chart” links point at QuantumLayers’ hosted ql-chart page, which renders the same visualization from nothing but URL parameters. Want a different cut, chasing by shift, a heatmap of behaviors, black squirrels by hectare? The data is public; go explore.

Data: 2018 Central Park Squirrel Census, via NYC Open Data, collected by The Squirrel Census. Counts reflect the dataset at request time.

This article is part of an ongoing series in which we analyze public datasets using QuantumLayers, our own data-analysis platform. Each post takes an openly available dataset, loads it into QuantumLayers, and shares what the charts and statistics reveal.